الفهرس 
Basic Concepts of Fuzzy SetsOnly 14 pages are availabe for public view
 |  | from | 132 |  |  |
| | | from | 132 | | |
Abstract
Fuzzy rule–based systems are developed based on fuzzy set theory, they consist of a set of IF-THEN rules. These rules are constructed by using qualitative or quantitative knowledge. Building a fuzzy rule–based system from a group of observed data is useful for simulation purposes and for use in controllers.
Construction of fuzzy rule-based system from input-output data has some interdependent sub-problems such as fuzzy partition of data space and identification of membership functions for the premises and consequences. In this thesis, the first problem is solved by fuzzy c-means clustering algorithm. The second problem is solved by Genetic Algorithm with fine tuning. The fuzzy c-means (FCM) clustering algorithm used to partition a collection of data points into a number of subgroups, where the points inside a cluster (a subgroup) show a certain degree of closeness or similarity. Due to its simplicity, and computational efficiency, it is a very popular technique Genetic Algorithms (GAs) are parallel, global search techniques that emulate natural genetic operators. GAs explore different areas of the parameter space, and then direct the search to regions where there is a high probability of finding improved performance. Because they simultaneously evaluate many points in the parameter space, they are more likely to converge toward the global solutions.
The main goal of this thesis is to apply GA-hybrid scheme to obtain fuzzy rule-based system with a small number of rules and high accuracy. To achieve this goal we will apply a procedure consisting of two parts: the first is called structure identification and the second part is called parameter identification. In the structure identification stage, FCM clustering algorithm have been applied to obtain the number of fuzzy partitions for the data space and the searching area of the parameters to identify. Every cluster that found by FCM clustering algorithm corresponds to one rule. The main advantage of using FCM for partitioning the data space is that, we do not have the exponential problem that occurs with using the ordinary fuzzy partition. In the parameter identification stage, the GA has been applied to obtain near optimal parameters for the fuzzy rule-based system. After GA search, optimal parameters for fuzzy rule-based system are finely tuned by gradient descent method. The advantage of using the combination of the GA with gradient descent method guarantees both global optimization and local convergence.